Know what AI says about you before you change a word.

What is an AI perception audit?

An AI perception audit is a structured reading of what AI engines say about your brand today. We ask ChatGPT, Gemini, Perplexity, Google AI Overviews, AI Mode and Claude the questions your buyers ask. Then we record how each one describes you, who it compares you to, who it recommends instead and which sources it trusts.

Most brands have never seen that story in full. Someone on the team checks one prompt on one engine, sees the brand name, and moves on. Or they don't see it, and assume the fix is more content.

An audit replaces that guesswork with evidence. It covers the whole buyer journey, not one query. It covers every engine that matters, not just the one your team happens to use. And it holds every answer up against the positioning you actually want.

It is not a ranking report

Rankings tell you where a page sits on a results page. AI answers work differently. There is no page two in AI search. An engine names a handful of brands, describes each in a sentence or two, and the buyer moves on.

So the question is not only whether you appear. It is what the engine says when you do. Are you described as the right fit for your best customers? Or as a cheaper alternative to someone else? Are the facts current? Is the category right?

That is why we call it a perception audit. We audit the description, the comparison and the reasoning behind them. A position number is only one small part of it.

Why does every engagement start with the audit?

Every engagement starts here because you can't change a story you haven't read. AI search is a positioning problem before it's a technical one. Engines recommend the brands they can describe in one sentence: who they're for, what they're best at, why they're different. The audit shows whether AI can write that sentence about you today, and what it writes instead.

Skip the audit and every later decision rests on guesses. You might rewrite pages that were never the problem. You might chase mentions on sites the engines don't read for your category. You might spend a quarter fixing the wrong engine.

There is an old idea in marketing that fits here. Ries and Trout called positioning the battle for the mind. The mind now includes an AI. Before you fight for it, you need to know what it already believes about you.

The audit is also the baseline

Every change we make later is measured against what we record here. When we sharpen your positioning or publish new proof, we rerun the same prompts and compare. Without a clean starting point, nobody can say what moved the answer.

Good marketers already work this way with ads and landing pages. You test against a control. In our work, the audit is the control. That is what we mean by message testing for AI.

What does the audit find out?

The audit answers six questions about your brand on every engine. What does AI say about you? Who does it compare you to? Which competitors does it recommend instead? Which sources does it trust and cite? Where is the story wrong or missing? And how far is it from the positioning you want? We record each answer per engine and per journey stage.

What AI says about you

We capture the words each engine uses to describe you. The category it puts you in. The customers it thinks you serve. The strengths and weaknesses it repeats. Small words matter here. "Affordable" and "enterprise-grade" send very different buyers your way.

Who it compares you to

Engines explain brands by comparison. We record which names appear next to yours, in what order and in what terms. Sometimes that comparison set is not the one your sales team expects.

Who it recommends instead

This is the painful part, and the most useful one. We note every prompt where a buyer who fits your ICP is sent to a competitor. We also note the reason the engine gives for that choice.

Which sources it trusts

Each engine leans on different sources. One quotes review sites. Another pulls from forums, comparison pages or your own docs. We map the domains and URLs each engine cites for your category, and mark the ones that mention you.

Where the story is wrong or missing

Outdated facts. A product line you retired. A feature you never had. Or silence on the thing you do best. We list each error with the prompt and the engine where it appeared, so it can be traced and fixed.

How far it sits from your positioning

Finally, we compare all of it with the story you want AI to tell. Where the two match, we protect it. Where they drift apart, we know exactly what to fix and where.

Why do we write the prompts in cold-buyer language?

We write audit prompts the way a buyer types before they have heard of you. If the prompts reuse your own positioning words, the engine echoes them back and the audit flatters you. Cold-buyer language shows the ground reality: what AI tells a stranger who simply has a problem. That is the version of the story your future customers will see.

This is the most common mistake in AI visibility measurement. A team builds its test prompts from its own website, sales deck or demo calls. The prompts ask about the exact features the brand is proud of. Unsurprisingly, the brand looks great.

Sales and demo language is excellent for copy. It is useless for measurement. It records how buyers talk after your team has framed the conversation. A cold buyer doesn't know your terms yet. They describe the problem in their own words, often short, messy and lowercase.

"We want to know the ground reality."

Krishna Kaanth M, Founder

What the difference looks like

Take a fictional brand, Acme, a data platform for RevOps teams. A prompt built from its own messaging might read: "best RevOps data platform with native CRM and billing joins". Acme wins that prompt easily, because the prompt is its own tagline.

A cold buyer types something closer to "our forecast keeps missing since we moved CRMs, what tools fix this". That buyer never says "data platform". Whether Acme shows up for that question is the real test, and it is the one the audit runs.

Our rules for audit prompts

• Two word banks, never mixed. Your language goes into copy. Buyer language goes into measurement.
• Criteria any buyer cares about. Does it work? Will my team use it? What happens when it breaks? Who do I call?
• The whole journey. Problem, solution, shortlist, compare, validate and decide.
• At least one scenario a competitor should win. It is the control, and it keeps the audit honest.
• Absence is a finding. We never tune the prompts until your brand shows up.

Single prompts and full conversations

Buyers rarely decide on one question. They open with a problem, add context over a few turns, let the engine build a shortlist, then question it. So alongside single prompts, we run short multi-turn conversations for your main ICP segments.

We never name your brand in those conversations. We watch where it enters, and where it drops out. Being named once is not winning. Still being named when the buyer asks "which one should I pick?" is winning.

How do we run the audit across engines?

We run the audit on two tracks. Max Visibility, our own AI search instrument, runs the prompt set on ChatGPT, Gemini, Perplexity, Google AI Overviews, Google AI Mode and Claude, several times each. Then a strategist reads every answer, because a score can't tell you what the words mean.

What Max Visibility records

Max Visibility is the instrument our team uses on every engagement. For the audit, we load your prompts, tag them by journey stage and ICP segment, and set up your competitor set. It then records:

• Every answer from each engine to each prompt, in full.
• Visibility, share of voice, average position and sentiment, for you and each competitor.
• A ranking per prompt, so we can see who gets named first.
• The domains and URLs each engine cites.
• Results by country, if you sell in more than one market.

The raw answers can be exported, so nothing lives only inside a chart.

Why several runs per prompt

AI answers are not fixed. Ask the same question twice and you can get a different list. One run is an anecdote. So each prompt runs several times on every engine, and we treat inclusion as a rate, not a yes or no.

Claude is in the audit too

We optimize for Claude on every engagement, so it belongs in the audit. Max Visibility tracks Claude as an add-on model, so its answers go into the same report as every other engine.

Why a strategist reads every answer

A dashboard can tell you how often you appeared. It can't tell you that three engines call you "a budget option" when you sell to the enterprise. That takes someone who knows your market and your intended positioning. Our team is brand strategists first, and reading the answers is the job.

On key prompts, we also ask the engine to explain its choice. Which criteria made a brand qualify? What kept others off the list? One explanation proves little, because a model's reasons are partly made up after the fact. Criteria that repeat across runs and engines are worth acting on.

What's in the perception report?

The perception report has five parts. You get the answers each engine gives across your buyer-journey prompts. You get your share of answer against competitors and a map of the sources each engine cites. You also get the gaps between AI's story and your intended positioning, plus a prioritized fix list. Every finding links back to the prompt, engine and answer it came from.

We walk your team through the report rather than sending a file into an inbox. Founders and marketing leaders often spot things only an insider would catch. An old product page, say, or a partner's outdated description. Those notes go straight onto the fix list.

Share of answer, explained

Share of answer is our plain name for how much of the AI conversation you own. In Max Visibility, it shows up as three numbers. Visibility is how often you are named. Share of voice is your share of all brand mentions. Position is how early you are named.

We read all three together. A brand can be named often but always last. Another can be named rarely but first. Those are different problems with different fixes.

How do we read what the audit finds?

We sort every gap by its cause, because each cause has a different fix. If engines don't retrieve you at all, it is an authority and content problem. If they cite your pages but leave you out of the answer, it is usually positioning. If they file you in the wrong category, the fix is re-anchoring. If you are included, the job is to protect it.

The second row surprises people most. A brand's own site can be among the most cited sources for its category, while the brand is left out of the recommendation. The engine trusts your content. It just can't tell what you are best at. That is a positioning fix, not a content fix.

A sample finding

Here is how a finding reads in the report, again for the fictional Acme. On shortlist prompts, Perplexity and ChatGPT cite Acme's own guides. Yet both recommend two rivals, and describe Acme as "an enrichment tool". Gemini doesn't mention it at all.

The diagnosis is "cited, but not named", plus a category problem. The fix list starts with the one sentence Acme wants AI to say. It then names the review profiles and comparison pages each engine leaned on, in the order to fix them.

Truthful fixes only

Every fix on the list is tagged by what is true today. Some strengths you have but under-communicate. Some need proof before you can claim them. Some are real product gaps, and we say so plainly.

We never recommend claiming something you can't back up. When ChatGPT names you, it stakes its own credibility on you. That is a trust transfer, and you can't buy your way into it.

What happens after the audit?

The audit feeds two pieces of work. Positioning and messaging takes the gaps and writes the story AI should tell, then makes it consistent everywhere AI looks. AI search experiments take the fix list and test each change against the audit baseline. Every result is measured on the same prompts, so you can see whether the story actually changed.

Into positioning and messaging

The perception gaps become the brief for positioning and messaging. We define who you are the right answer for and the one sentence AI should say about you. Then we write the claims, proof points and objection answers behind it, and make your site, reviews, docs and press say the same thing.

Into AI search experiments

Each fix becomes a hypothesis for AI search experiments. For example: if your site, reviews and press all say you are built for one segment, engines will stop calling you a general tool. We record the change, rerun the baseline prompts and compare. We keep what clears the noise band and drop what doesn't.

Rerun on a schedule

AI answers drift as models update and new sources appear. So the audit is not a one-off. We rerun the same prompt set on a regular cadence, usually monthly, and keep the prompts stable so results stay comparable.

That turns the first audit into a trend line. You see which engines moved, which sources changed and which competitors gained ground. It is also how you prove the work is paying off.

Into GEO

Some gaps are retrieval or authority problems. Those go to our GEO work: answer-ready content, SEO foundation, earned citations and PR, video and social. The source map tells us where to publish first.

Who is the AI perception audit for?

The audit is for B2B SaaS and AI companies, roughly $5M to $100M ARR, whose buyers now research with AI. It suits founders, CMOs, VPs of Marketing and Heads of Growth who answer for pipeline and want to know what AI tells their buyers. It is not for teams who want results this week, or who want paid media.

Signs it is time

• Prospects on sales calls repeat something about you that isn't true.
• You changed your positioning, category or product line, and AI hasn't caught up.
• Competitors show up in AI answers for questions you should own.
• Your CEO or board asks "what does ChatGPT say about us?" and nobody has a full answer.
• You rank well on Google, but very few leads say they found you through AI.

When it is not the right fit

If you want results this week, a keyword list with a new name, or paid ads, we are not the right team. We are inbound only. We earn your place in AI answers. We never buy it.

If you want to see the story AI tells about you today, book a GEO strategy call.